Re: [PATCH v4 08/12] phy: phy-mtk-dp: Add support for digital and analog calibration

Thank you for your contribution! Sashiko AI review found 3 potential issue(s) to consider:
- [Medium] Silent ignoring of hardware write failures during PHY initialization.
- [Medium] False positive error logs emitted when optional eFuse calibration data is missing.
- [Low] Unused variables left over from refactoring (`dev` and `ret`).
--

commit 7b4b9550381f2fc0574d6e28361a57bed4016c26
Author: AngeloGioacchino Del Regno [off-list ref]

phy: phy-mtk-dp: Add support for digital and analog calibration

This commit adds support for reading DP PHY calibration values from eFuse.
If calibration values are present, they are written to the PHY; otherwise,
the driver falls back to default SoC-specific values. It also adds support
for writing the calibration values for the analog part of the PHY.

[ ... ]

-static int mtk_dp_phy_init(struct phy *phy)
+static void mtk_dp_phy_set_digital_drv_params(struct mtk_dp_phy *dp_phy)
 {
-	struct mtk_dp_phy *dp_phy = phy_get_drvdata(phy);
 	const struct mtk_dp_phy_pdata *pdata = dp_phy->pdata;
 	const u32 reg = pdata->regs_dig_lane[DP_PHYD_LAN_DRIVING_PARAM_0];
-	static const u32 driving_params[] = {

[ ... ]

-	};
 	int i, ret;

[Severity: Low]
Is `ret` still needed here? The return checks for regmap bulk writes were
removed below, which appears to leave this variable completely unused.

 
 	/*

[ ... ]

-	for (i = 0; i < MTK_DP_PHY_MAX_LANES; i++) {
-		ret = regmap_bulk_write(dp_phy->regmap,
-					pdata->off_dig_lane[i] + reg,
-					driving_params,
-					ARRAY_SIZE(driving_params));
-		if (ret)
-			return ret;
-	};
+	for (i = 0; i < MTK_DP_PHY_MAX_LANES; i++)
+		regmap_bulk_write(dp_phy->regmap, pdata->off_dig_lane[i] + reg,
+				  pdata->driving_params,
+				  ARRAY_SIZE(pdata->driving_params));
+}

[Severity: Medium]
Does this code introduce a regression by silently ignoring hardware write
failures during PHY initialization?

Prior to this patch, if regmap_bulk_write() failed, the error was propagated
back to the caller. Now that this has been extracted into a void function,
any hardware or bus errors during these bulk writes will be silently dropped.

+
+static int mtk_dp_phy_init(struct phy *phy)
+{
+	struct mtk_dp_phy *dp_phy = phy_get_drvdata(phy);
+	struct device *dev = &phy->dev;

[Severity: Low]
Is the `dev` variable used anywhere in this function? It appears to be
declared and assigned but never utilized.

+
+	mtk_dp_phy_set_digital_drv_params(dp_phy);
+	mtk_dp_phy_set_analog_calibration_params(dp_phy);
 
 	return 0;
 }

[Severity: Medium]
Does the PHY initialization now unconditionally succeed regardless of
the underlying hardware state?

Because the extracted functions mtk_dp_phy_set_digital_drv_params() and
mtk_dp_phy_set_analog_calibration_params() return void, any failures
during regmap updates will not be caught, and this function will always
improperly return 0 on failure.

[ ... ]

+static int mtk_dp_phy_get_one_cal_para(struct device *dev, const char *name, u8 max_val)
+{
+	u8 buf_byte;
+	u16 buf;
+	int ret;
+

[ ... ]

+	ret = nvmem_cell_read_u8(dev, name, &buf_byte);
+	if (ret)
+		ret = nvmem_cell_read_u16(dev, name, &buf);
+	else
+		buf = buf_byte;
+
+	if (ret)
+		return dev_err_probe(dev, ret, "Cannot get calibration data for %s\n", name);

[Severity: Medium]
Will this cause false positive error logs on normal boots?

If an NVMEM cell is missing, the nvmem read functions naturally return
-ENOENT. This dev_err_probe() triggers a KERN_ERR log here before returning
the error to the caller.

+
+	if (buf == 0) {

[ ... ]

+static int mtk_dp_phy_get_calibration_data(struct mtk_dp_phy *dp_phy)
+{

[ ... ]

+end:
+	if (ret < 0) {
+		/*
+		 * If any of the calibration values is missing, or if there
+		 * is no calibration at all in the eFuses, copy the default
+		 * one entirely (as partial values shall not be mixed!)
+		 */
+		if (ret == -ENOENT) {
+			dev_info(dev, "Using calibration default values\n");
+			mtk_dp_phy_get_default_cal_data(dp_phy);
+			return 0;
+		}

[Severity: Medium]
Since -ENOENT is explicitly checked and handled gracefully as a benign
fallback here, should we avoid logging it as an error in
mtk_dp_phy_get_one_cal_para() earlier?

Logging an expected missing cell as an error falsely spams the kernel dmesg
with an error trace for boards that legitimately omit calibration data.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260910092038.48291-1-angelogioacchino.delregno@collabora.com?part=8

-- 
linux-phy mailing list
linux-phy@lists.infradead.org
https://lists.infradead.org/mailman/listinfo/linux-phy